REBECCA L. LODA-HUTCHINSON

Transcription

1 ANALYSIS OF NF1 MUTATION MECHANISMS By REBECCA L. LODA-HUTCHINSON A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA

2 2009 Rebecca L. Loda-Hutchinson 2

3 To Mrs. Michelle Doyle and Mr. Stephen Sans, whose inspiration and enthusiasm started me on this path many years ago 3

4 ACKNOWLEDGMENTS First, I need to thank all the families and individuals who participated in these studies; without them, this work would not have been possible. I want to thank Dr. Peggy Wallace for taking me on as a student and letting me work in her lab for the past five years. I would not have made it through this experience without her guidance, understanding, and support. I also want to thank the members of my supervisory committee, Dr. Daniel Driscoll, Dr. Keith Robertson, and Dr. Peter Sayeski, for their time, support, and suggestions throughout this process. My thanks also goes to all the past and present members of the Wallace lab, especially Beth Fisher, for her patience and smiles as I was learning the ropes; and Michelle Burch, for always having an answer for me. I also want to thank Dr. Michele Tennant, Dr. Pauline Ng (Fred Hutchinson Cancer Research Center, Seattle), and Maya Zuhl (University of Maryland Biotechnology Institute) for their invaluable bioinformatics help. Finally, I want to thank my family and friends for their support and encouragement no matter what has come my way these past five years. I especially want to thank Randi Marie, Deborah, Nicole, and Lauren, who have shared in the graduate school experience with me, for helping me stay sane, giving me someone to talk to, and reminding me to relax and have fun. Last, but not least, a huge thanks goes to my husband, Lance, who has seen this process through with me from applications to a dissertation, and never stopped believing that I could do it. 4

5 TABLE OF CONTENTS ACKNOWLEDGMENTS...4 LIST OF TABLES...7 LIST OF FIGURES...8 ABSTRACT...9 CHAPTERS 1 INTRODUCTION...11 page Neurofibromin and the NF1 Gene...11 Neurofibromatosis 1 (NF1)...12 Mutations in NF SOMATIC CpG C TO T TRANSITIONS AT NF1 GERMLINE HOTSPOTS...19 Introduction...19 Materials and Methods...22 Mutation Detection by PCR and Restriction Digest...22 Methylation Status Analysis...23 Results...24 Methylation Status Analysis...24 Mutation Detection by PCR and Restriction Digest...26 Discussion ALTERNATIVE SPLICING OF EXON 23a AND mrna EDITING...32 Introduction...32 Material and Methods...36 Reverse-Transcription and PCR...36 Cloning and Sequence Analysis...37 Results...37 Alternative Splicing Patterns...37 Analysis of RNA Editing...39 Discussion MISSENSE MUTATION COMPUTATIONAL ANALYSIS OF PATHOGENICITY...47 Introduction...47 Materials and Methods...49 Missense Computational Methods...49 Other Databases and Programs

6 Data Sets...50 Sequence Input Requirements...50 Splice Analysis...53 Results...53 Discussion CONCLUSIONS AND FUTURE DIRECTIONS...64 Somatic CpG C to T Mutation...64 Alternative Splicing of exon23a...66 Computational Analysis Comparison...69 APPENDIX A CpG C to T Mutation Analysis Data...70 B Exon 23a Alternative Splicing Data...74 C Missense Mutation Computational Analysis Data...77 LIST OF REFERENCES...85 BIOGRAPHICAL SKETCH

7 LIST OF TABLES Table page 2-1 Primers used in CpG C to T mutation screening Summary of alternative splicing of exon23a seen in various sample types examined Summary of computational missense prediction results for 5 data sets (4 controls and 1 unknown) A-1 Results of CpG C to T mutation screen using TaqαI restriction enzyme digest...70 B-1 Relative concentrations of Type I v Type II mrna in blood, tumor and culture samples...74 C-1 Results for each mutation as returned by the various computational methods used

8 LIST OF FIGURES Figure page 1-1. Diagram of neurofibromin s putative domains Methylation-sensitive restriction digest protocol Visualization of methylation-specific restriction digest analysis Results of bisulfite sequencing TaqαI digests of E23.2 PCR products visualized on 8% PAGE Representative gels showing relative concentrations of Type I v Type II mrna in various tissue types studied Comparison of alternative splicing of exon23a in primary plexiform tumors v cultured Schwann cells from the same tumors

9 Chair: Margaret R. Wallace Major: Medical Sciences Genetics Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ANALYSIS OF NF1 MUTATION MECHANISMS By Rebecca L. Loda-Hutchinson May 2009 Neurofibromin is a large protein encoded by the NF1 gene, whose best understood function is as a down-regulator of the RAS signaling pathway, leading to NF1 s classification as a tumor suppressor gene. NF1 gene mutations, which occur at a rate 10x higher then the genome average, lead to the autosomal dominant disorder neurofibromatosis 1 (NF1). NF1 has variable expressivity and is clinically diagnosed using seven diagnostic criteria, of which the key features are café-au-lait spots, neurofibromas, Lisch nodules, and skin fold freckling. Genetic diagnosis is difficult, as the gene is very large and very few mutations are recurrent. Additionally, many NF1 phenotypes are believed to originate with a second mutation in the wild type allele in specific cell types. To gain insight into mutation mechanisms in NF1, I pursued three projects. First, I analyzed the rate of somatic C to T mutations at four hotspots for germline mutations. The methylation status of these sites in somatic cells makes them susceptible to C to T transitions; however no such mutations were identified in 123 neurofibromas. Next, the alternative splicing of exon23a, the inclusion of which reduces neurofibromin s RAS-GAP function, was examined in various tissue types and tumors. Transcripts containing exon23a (Type II mrna) are predominant in most tumors; mrna lacking exon23a (Type I) is predominant in blood leukocytes. While previously reported in tumors containing increased 9

10 Type II mrna, no RNA editing was observed in the tumors in this study. Finally, I tested the accuracy of computational methods at predicting the effects of NF1 missense mutations (pathogenic versus neutral). These programs are needed clinically since mutations can not be tested functionally. No program was 100% accurate, but each had advantages in different situations. This work contributes to the knowledge in NF1, toward a goal of targeted therapies and improved diagnosis. 10

11 CHAPTER 1 INTRODUCTION Neurofibromin and the NF1 Gene Neurofibromin is a 220 kd, 2818 amino acid protein expressed in vertebrate and non-vertebrate animal species, with two homologs in yeast. While it is ubiquitously expressed, it is found at highest levels in Schwann cells, neurons, and other neural crest-derived cell types (Daston et al., 1992). Neurofibromin is known to be localized to the cytoplasm and interacts with microtubules (Gregory et al., 1993), but recent work suggests that it may also travel to the nucleus (Vandenbroucke et al., 2004) and may localize to different cell compartments based on stage of development and cell type (Roudebush et al., 1997; Kaufmann et al., 2002; Malminen et al., 2002; De Schepper et al., 2006). Additionally, four isoforms caused by alternative splicing have been shown to occur at various developmental stages as well as in specific cell types, which will be discussed in more detail in Chapter 3 (Nishi et al., 1991; Suzuki et al., 1992; Gutmann et al., 1993a; reviewed by Skuse and Cappione, 1997). Figure 1-1 shows a diagram of the putative domains of neurofibromin. While several functional domains have been proposed, including a microtubule binding domain (Gregory et al., 1993) and a SEC14-like domain (D Angelo et al., 2006), the most well-characterized functional domain of neurofibromin is a 360 amino acid RAS-GTPase activating protein (GAP)-related domain (GRD) (Ballester et al., 1990, Xu et al., 1990; reviewed by Cichowski and Jacks, 2001). This domain binds activated p21-ras-gtp and leads to the hydrolysis of GTP to GDP, inactivating the RAS protein. This inactivation decreases downstream RAS signaling, and thereby down-regulates cell proliferation and the inhibition of apoptosis. Since increased RAS signaling is associated with tumorigenesis, neurofibromin s inhibitory activity, as well as its gene s adherence to Knudson s two-hit hypothesis (Knudson, 1971), led to its classification as a tumor suppressor (Colman et al., 1995). 11

12 It is not surprising then, that mutations in the gene encoding neurofibromin predispose individuals to developing multiple tumor types. Neurofibromin is encoded by the NF1 gene, which contains 60 exons that span 280 kb of DNA on chromosome 17q11.2 (Cawthon et al. 1990; Viskochil et al. 1990; Wallace et al. 1990; Li et al. 1995). The GRD is encoded by exons 21 to 27a, and mutations in this region have been studied functionally in yeast. Expressing this domain alone in yeast lacking the yeast NF1 homologs (Ira1, Ira2) can restore a normal phenotype (Ballester et al., 1990). While the GRD is the best understood domain, mutations have been found in all exons/domains of the NF1 gene, which lead to neurofibromatosis 1 (NF1). Neurofibromatosis 1 (NF1) Although the reasons are unclear, the mutation rate at the NF1 locus is ten times higher than the average rate in the human genome, making NF1 one of the most common genetic disorders (Riccardi, 1992; Hughes, 1994). It is inherited in an autosomal dominant manner, and occurs in approximately 1 in 3,500 births worldwide (Riccardi, 1992). Occurrence rates do not vary based on sex or ethnicity. Approximately half of all cases brought to medical attention have no affected parent and represent a de novo NF1 mutation. Of these de novo mutations, 80% of those that are not deletions are paternal in origin (Stephens et al., 1992). The physical manifestations of NF1 are numerous and vary greatly in severity among individuals (variable expressivity) (Carey and Viskochil, 1999). In addition, even within a family, phenotypic features and severity may vary, although there is less intra-familial variation than inter-familial (Easton et al., 1993; Szudek et al., 2002). Because of this, seven diagnostic criteria were established in 1988 by the National Institutes of Health to aid in clinical diagnosis (reviewed by Gutmann et al., 1997). These seven diagnostic criteria are: six or more café-au-lait macules (must be 0.5 cm or larger before puberty, 1.5 cm or larger after puberty), neurofibromas (benign Schwann cell tumors), optic pathway tumors (benign pilocytic astrocytomas), two or more Lisch 12

13 nodules (benign hyperpigmented lesions of the iris that look like freckles), skeletal dysplasia (typically sphenoid bone dysplasia or tibial dysplasia), skin fold freckling, and having a first degree relative with NF1. A patient is clinically diagnosed with NF1 if they meet any two or more of these criteria. Although molecular genetic testing is available, these criteria remain important in diagnosis as many patients do not have the expensive genetic analysis done (McClatchey, 2007). NF1 is a progressive disorder, and these phenotypic features may arise or increase unpredictably over time. Café-au-lait macules are often the earliest phenotype to develop, usually present by 2 years of age. Tibial dysplasia and skin fold freckling develop in the first few years of life as well, with the majority of other phenotypes arising by late childhood/early adolescence (Riccardi, 1992; Williams et al., 2009). In addition to these criteria, patients with NF1 are also at increased risk for additional complications, including learning disabilities, short stature, scoliosis, renal artery stenosis, hypertension, macrocephaly, and increased risk of certain malignancies, such as rhabdomyosarcoma, pheochromocytoma, and myeloid leukemia (reviewed by Gutmann et al., 1997). Of the main phenotypic characteristics, neurofibromas generally cause the most trouble and discomfort to the patients. Discomfort may include itching, pain, and tenderness around areas of tumor growth, and furthermore patients can suffer socially due to disfigurement (Riccardi, 1992; Gottfried et al., 2006). Neurofibromas originate from the peripheral nerve Schwann cells and are clonal in origin (Skuse et al., 1991; Colman et al., 1995). Schwann cells are one of several cell types that make up the peripheral nerve sheath, which surrounds the nerve and axons and helps maintain proper nerve function. The healthy peripheral nerve sheath is well organized, with Schwann cells closely associated with the axon of the nerve, producing the myelin used to insulate the axon. Disorganization and an increase in all peripheral nerve sheath 13

14 cell types are seen in neurofibromas (reviewed by Cichowski and Jacks, 2001). Thus, these tumors contain primarily clonal, expanded (and somatically mutated, as described below) Schwann cells, but also contain other cell types normally associated with peripheral nerves, such as non-clonal Schwann cells, fibroblasts, mast cells, and axons (Serra et al., 2000). The cutaneous, or dermal type neurofibromas, are derived from Schwann cells associated with nerve twigs, close to or at the surface of the skin. Some patients develop thousands of these tumors while others have few or none. They rarely grow beyond one centimeter in diameter and are not known to become malignant. In contrast, plexiform neurofibromas develop from Schwann cells associated with larger nerves, are generally deeper in the body, are larger then dermal neurofibromas, and may involve large areas of the body. It is estimated that 30-40% of NF1 patients develop plexiform tumors, some of which may be asymptomatic nodular masses in the thorax (Tonsgard et al., 1998). In addition to the risk that a plexiform tumor could grow quite large and disrupt bone and organs, 10-15% of plexiform neurofibromas will transform into malignant peripheral nerve sheath tumors (MPNSTs), presumably by the accumulation of genetic and epigenetic changes (Ferner and Gutmann, 2002). Unfortunately, MPNSTs often become metastatic and have a poor prognosis unless removed completely by surgery. How the transformation from benign plexiform neurofibroma to MPNST occurs is unclear. The development of benign neurofibromas is somewhat better understood and the initiating event appears to follow Knudson s two-hit hypothesis of tumor suppressors (1971). Because of the dominant nature of NF1, the majority of patients are constitutionally heterozygous for an inactivating mutation in the NF1 gene. In a minority of patients the disease-causing mutation arose at a point in development shortly after the single cell stage, leading to mosaic distribution of the mutation (Colman et al. 1996). Regardless, neurofibromas arise when Schwann cells 14

15 carrying this mutated NF1 allele develop a second mutation that reduces the function of the previously wild-type NF1 allele (Serra et al., 1997, 2000; reviewed in MacCollin and Wallace, 2002). When a second, somatic mutation or other inactivating event interferes with the function of the wild-type NF1 allele in a Schwann cell, neurofibroma development is thought to start or at least be poised to begin upon additional signals. This somatically-mutated Schwann cell then undergoes clonal expansion, dividing inappropriately, even without axonal contact. Similarly, optic pathway tumors initiate from two NF1 mutations in an astrocyte (Bajenaru et al., 2003), tibial dysplasia is associated with two mutations in osteoblasts (Stevenson et al., 2006), and café-au-lait macules originate from a melanocyte with two NF1 mutations (DeSchepper et al., 2007). Mutations in NF1 The NF1 germline mutation rate is estimated between 1/7,800 and 1/23,000, approximately 10-fold higher than average (reviewed in Gottfried et al. 2006). It is unclear why the NF1 locus is so susceptible to mutation; the size of the gene alone does not explain this phenomenon (Friedman 1999; Fahsold et al; 2000). The NF1 germline mutation spectrum is broad, with over 1000 germline mutations identified. Of these, 70-80% are clearly inactivating (frameshift and nonsense mutations), and none recur in more than 2% of cases (reviewed in Thomson et al. 2002). Approximately 20% are aberrant splicing mutations (Messiaen et al., 2000; Serra et al., 2001; Wimmer et al., 2007; Pros et al., 2008). The exception to this is a Mb microdeletion of the region of chromosome 17 that contains the NF1 gene and 14 flanking genes. This microdeletion was first identified in patients with an early age of tumor onset, high tumor loads, specific facial features, and mental retardation (Kayes et al., 1992, 1994; Leppig et al., 1996, 1997; Wu et al., 1995, 1997). Later studies, however, revealed patients with no outstanding phenotype who also carry microdeletions in the NF1 region (Rasmussen et al., 1998; 15

16 Dorschner et al., 2000). This microdeletion is thought to account for 4-7% of NF1 cases, and approximately 80% of de novo microdeletions are found on the maternally inherited chromosome 17 (Lázaro et al., 1996; Valero et al., 1997; Rasmussen et al., 1998; Upadhyaya et al., 1998; Lopez-Correa et al., 2001; Kluwe et al., 2004). These deletions are mediated by unequal crossover meiotic events in the germline at mis-aligned repetitive regions. Both the 1.2 and 1.4 Mb versions of the microdeletion share a 5 repeat sequence containing a pseudogene of JJAZ1, called JAZFP (Forbes et al., 2004). The 1.2 Mb deletion s 3 end is in the JJAZ1 gene, whereas the 1.4 Mb deletion s breakpoint is distal to JJAZ1 (Raedt et al., 2006). In addition to the NF1 gene being deleted, 14 other flanking genes are deleted as well, leaving patients hemizygous for these genes. These deletions can also occur somatically, early in embryogenesis (mitotic recombination), resulting in a patient who is mosaic for an NF1 microdeletion (Rasmussen et al., 1998; Kehrer-Sawatzki et al., 2004) and who may show fewer NF1 features. A much smaller body of knowledge exists for somatic NF1 mutations. Loss of heterozygosity was first seen in neurofibromas by Colman et al. (1995). Since then, it has become clear that in most cases the allelic imbalance is due to mitotic recombination, where a region of the mutant chromosome replaces that of the wild-type chromosome (Serra et al., 2001a). Loss of heterozygosity has been found in 10-40% of tumors analyzed (Rasmussen et al., 2000; Serra et al., 2001; Upadhyaya et al., 2004). Sawada et al. (1996) were the first to identify a specific somatic mutation in the NF1 gene in a tumor from a patient whose germline mutation was already known. Since then reports of additional somatic mutations have been limited, and there is still little known about the somatic mutation spectrum (Rasmussen and Wallace, 1998; Upadhyaya and Cooper, 1998; Eisenbarth et al., 2000; Wiest et al., 2003; Upadhyaya et al., 2004). The difficulty of identifying these mutations is due not only to the size and complexity of 16

17 the NF1 gene, but also to the fact that each tumor is thought to arise from an independent inactivating event, so that somatic mutations differ not only between individuals, but also between tumors from the same individual (Colman et al., 1995; Serra et al., 2001b). Furthermore, the tumors have an admixture of cells and thus only a portion of the cells (the clonally-expanded Schwann cells) carry the somatic mutation. Further understanding of the nature and frequency of somatic mutations will provide important information about risk factors, disease progression and tumorigenesis, as well as help elucidate the pathways that lead to neurofibroma formation and the transformation to MPNSTs. This work adds to this body of knowledge by addressing three specific areas of investigation directly related to the somatic mutation spectrum of NF1. Chapter 2 discusses the somatic rate of C to T transitions at four sites known to be hotspots for this mutation type in the germline. Chapter 3 examines alternative splicing of exon23a in NF1-related tumors, its expression level in various cell types, and its relationship to RNA editing. Chapter 4 addresses the need for reliable ways to predict the effects of missense mutations (both somatic and germline) on neurofibromin function, and evaluates several computational methods that make such predictions. 17

18 Figure 1-1. Diagram of neurofibromin s putative domains. Alternatively spliced exons are shown, along with the number of amino acids they insert. Well-characterized domains, with known crystal structure, are shown in red. GRD = GAP-related domain (exons21-26); Sec14 = SEC14 protein-like domain (exons27b-28). Domains shown in blue are not as well-characterized. TM = Transmembrane domain (exon10a2); CSRD = Cystein/serine-rich domain (amino acids ); TBD = Tubulin-binding domain (amino acids ); SB = Syndecan-binding domains (amino acids (with in the GRD), ); LRD = Leucine-rich domain (Sec14 domain is within the LRD) (amino acids ); NLS = Nuclear localization sequence (exon43). 18

19 CHAPTER 2 SOMATIC CpG C TO T TRANSITIONS AT NF1 GERMLINE HOTSPOTS Introduction In mammalian DNA, a cytosine immediately 5 to a guanine, designated CpG, can be methylated, having a methyl group attached to carbon 5, producing 5-methylcytosine (5mC). This is the only nucleotide that can be methylated in mammalian DNA (Bird, 2002). Cytosines not in the context of CpG are not subject to methylation, with the exception of those found in a CpNpG context (Clark et al., 1995), and it is believed that the majority of CpG dinucleotides not in CpG islands are methylated (Cooper and Krawczak, 1993). 5-methylcytosine can be associated with transcriptionally inactive regions, especially if such methylation occurs in a gene promoter region where CpGs can be clustered (termed a CpG island) (Bird, 1986) or in other transcriptional regulatory elements (Cooper and Krawczak, 1993). Cytosine residues in CpG dinucleotides are susceptible to the spontaneous loss of the amine group at carbon 4 by hydrophilic attack or by chemical deamination. Once the amine group is lost, a tautomeric shift can occur. In the unmethylated state this shift produces uracil, while 5mC becomes thymine (a C to T transition). There are mismatch repair enzymes to recognize and repair both G:U and G:T mispairs, but G:U mispairing is more efficiently repaired, in part because uracil is not used in the production of DNA (Cooper and Krawczak, 1993; Lari et al., 2002, 2004). Brown and Jiricny (1987) found that G:T mispairs (resulting from C to T transitions) were repaired in favor of guanine 90% of the time. These mispairs were not repaired 2% of the time, and in the remaining 8% of cases, they were corrected in favor of thymine, causing the C to T transitions to become fixed in the cell s DNA. These base changes can be neutral (e.g. in non-coding DNA, or a silent coding-region substitution), or they can cause errors in splicing or coding regions leading to a premature stop codon, amino acid deletion, or amino acid substitution. No neutral CpG C to T 19

20 transitions have been reported in the NF1 gene despite studies that would have detected such changes in the coding regions and UTRs; no such polymorphisms have been found in the 11 Kb mrna. When considering 880 base changes previously reported to be involved in genetic disorders, Cooper and Krawczak (1993) found that 32.8% were C to T or G to A (resulting from a C to T transition on the antisense strand) mutations at CpG dinucleotides. One such mutation has been repeatedly identified as a germline mutation in unrelated neurofibromatosis 1 (NF1) patients, accounting for 1-2% of mutant alleles. This CpG C to T mutation, R1947X (CGA to TGA), results in a stop codon in exon31 of the NF1 gene (Horiuchi et al., 1994; Lazaro et al., 1995; Dublin et al., 1995). There are several other such nonsense germline mutations due to CpG C to T transitions, each of which has a frequency of 0.5-2%: R416X, R440X, R816X, R1241X, R1276X, R1362X, R1748X, and R2429X. C to T transitions account for 18-30% of NF1 germline mutations, most causing nonsense codons, the rest missense mutations (Krkljus et al., 1998; Fashold et al., 2000; Messiaen et al., 2000). Of these, the majority that were de novo arose in the paternal genome (Jadayel et al. 1990; Wallace et al. 1991; Krkljus et al. 1998), which may be related to methylation during spermatogenesis (Driscoll and Migeon, 1990). In contrast, point mutations at CpG sites account for approximately 50% of the reported germline mutations in TP53 (Greenblatt et al., 1994), ~30% reported for RB1 (Lohmann et al., 1996), and 54% of germline mutations in NF2 (Baser, 2006). These mutations may be under-represented in NF1 relative to some other genetic disorders. In neurofibromatosis 2 (NF2), caused by mutations in the NF2 gene, CpG C to T mutations are reported to account for 38-52% of somatic mutations (Baser 2006). While there is a strong body of knowledge on the germline frequency of these mutations in NF1, there are no studies focused on the rate at which they occur somatically. There are multiple mutation screens that 20

21 have identified somatic NF1 CpG mutations (Sawada et al., 1996; Eisenbarth et al., 2000; Serra et al., 2001b; Wiest et al., 2003; Upadhyaya et al., 2004, 2008; Maertens et al., 2006; Bottillo et al., 2009). In many of these studies, mutations were not identified in more then half the samples. Only 142 mutations of any type were identified out of 372 tumors. These studies screened for mutations in germline C to T transition hotspots, such as the previously mentioned R1947X mutation. Of the NF1 somatic mutations identified, 7/142 (4.9%) were CpG mutations. Of the 7 CpG somatic mutations found, 5 were at a previously identified germline mutation site (Eisenbarth et al., 2000; Wiest et al., 2003; Upadhyaya et al., 2004, 2008; Bottillo et al., 2009). While NF1 contains more CpG containing codons, and specifically CGA codons (which create a premature stop codon with a C to T transition) than NF2, it does not appear to be as susceptible to CpG C to T mutations. Further identification of somatic mutation mechanisms is important since somatic inactivation of the wild-type NF1 allele in NF1 patients is the initiating step in neurofibroma tumorigenesis in Schwann cells. It is also possible that sporadic neurofibromas are the result of two somatic NF1 mutations in a single cell; one such case has been reported, involving chromosome translocation (Storlazzi et al., 2005). The goal of my work was to test whether somatic C to T transitions at CpG dinucleotides in the NF1 gene may be a common mechanism of generating a second hit in Schwann cells. Furthermore, tumors that have any DNA hypermethylation (which may be present at low levels in some plexiform neurofibromas, based on promoter studies by Fishbein et al. (2005)) could be at risk for an increase in CpG transitions since more cytosines are methylated. Further knowledge of the types and frequencies of somatic mutations such as C to T transitions will be useful for understanding genetic changes 21

22 contributing to NF1 tumors. In this work, I will test for somatic presence of each of four recurrent germline NF1 C to T transition stop mutations in tumor DNA. Materials and Methods Mutation Detection by PCR and Restriction Digest One hundred ninety-seven DNAs were previously prepared by phenol/chloroform extraction from NF1 patient blood and tumor samples as described in Colman et al. (1993) and kept in concentrated stocks at 4 C. PCR primers were designed previously for exon10a, 22, 23.2, and 41 of the NF1 gene. These exons contain the mutations R440X, R1241X, R1362X, and R2429X, respectively. The primers are in the introns flanking each exon, and details are given in Table 2-1. Diluted DNA samples were used as template ( ng) in polymerase chain reactions (PCR) under standard conditions. These exon PCR products were then digested using the TaqαI restriction enzyme in 25 µl reactions incubated at 65 C for 2 hours, with more enzyme added after 1 hr. The sequence recognized and cut by TaqαI spans the CpG site of interest in each exon, and any changes in the sequence will result in failure of the enzyme to cut. These digest reactions were visualized using ethidium bromide staining after electrophoretic separation on 8% native polyacrylamide gels. With this assay, germline heterozygotes for one of these mutations show an uncut band plus two smaller bands in the gel. The presence of an uncut fragment was suggestive of the presence of a mutation and a second digest was used to confirm the presence of uncut DNA, and controls were used to ensure complete digestion. Table 2-1 lists the size of both the uncut fragments and the cut fragments produced by TaqαI and methylsensitive digestion for each exon. Numbers of mutation-positive and mutation-negative samples at each site were determined, as well as total mutations per tumor type. 22

23 Methylation Status Analysis In addition to screening for the four mutations, I also investigated whether those cytosines were methylated in normal somatic Schwann cells, to ensure that they could be a target for 5meC deamination. This was first done using methyl-sensitive restriction digest, and then bisulfite genomic sequencing. A schematic of the methyl-sensitive restriction digest method is given in Figure 2-1. Genomic DNA was used in restriction digest reactions using methylation-sensitive restriction enzymes (BstBI for sites in exon10a, 23.2 and 41, incubated at 65 C; AvaI for exon22, incubated at 37 C). These enzymes recognize and cut the sequence spanning the CpG sites of interest, but cut only if the cytosine residues are unmethylated. The sequence surrounding the CpG site of interest in exon10a is 5 GGTTGAACTTCGAAATATGTTT 3 ; in exon41 the sequence is 5 TGAAGAAGTTCGAAGTCGCTGC 3 ; in exon23.2 the sequence is 5 CCCTCAACTTCGAAGTGTGTGC 3. These three sites are cut by BstBI, which recognizes and cuts the sequence 5 TT/CGAA 3. The sequence surrounding the CpG site of interest in exon22 is 5 TGAACTAGCTCGAGTTCTGGTT 3. This site contains the recognition sequence of AvaI, which is 5 C/YCGRG 3, where Y is either T or C, and R is either A or G. PCR reactions were then set up using DNA digested with these enzymes as well as untreated DNA. The primers listed in Table 2-1 were used in these reactions. This same methylation analysis was carried out on control samples as well as samples from a plexiform neurofibroma. As a control, PCR samples were set up using the same samples, and these reactions were digested with TaqαI restriction enzyme as described above. The TaqαI site (TC/GA) is within the recognition sites of the methyl-sensitive enzymes, and is cut by TaqαI whether the C of interest is methylated or not. The presence of digested product in these reactions will confirm that the recognition sites are not mutated in the samples tested, and are therefore able to be digested by 23

24 the methyl-sensitive enzymes. All digestion products were visualized using ethidium bromide staining after electrophoretic separation on 8% native polyacrylamide gels. The CpG site in exon23.2 (R1362X) was chosen for analysis by bisulfite sequencing as an independent measure of somatic methylation. Genomic DNA from cultured normal human Schwann cells was subjected to sodium bisulfite treatment per the method given in Fishbein et al. (2005). The treated DNA was then PCR amplified using primers 23-2MEF (5 GTTAGAATTATTAGAGAGTTTTGAG 3 ) and 23-2MER (5 ATAATCTCTAACTATAAACATACCTAATA 3 ) with an annealing temperature of 54 C. The sequence of these primers was determined assuming the conversion of unmethylated cytosine to thymine after sodium bisulfite treatment. These primers amplify a 158 bp fragment spanning exon23-2 and 23 bp of intron23-2. The PCR products were ligated into a plasmid vector (Topo TA, Invitrogen) and transformed into chemically competent E. coli cells (One Shot TOP10, Invitrogen) using the manufacturer s protocol. The cells were plated on LB-ampicillin agar plates with IPTG and X-gal, and incubated at 37 C overnight. Bacterial colonies (clones) positive for the insert were identified by blue/white selection, followed by PCR amplification of picked colonies using the original primers. The PCR products from five of these clones were sequenced, using cycle sequencing with the ABI Prism Big Dye 3 system and the UF CEG sequencing core, with the 23-2MEF primer as the sequencing primer. Results Methylation Status Analysis In one set of experiments, the methylation status of all four CpG C to T germline mutation hotspots was analyzed using methylation-sensitive restriction digest analysis. I examined these sites in a normal Schwann cell culture, leukocytes from two non-nf1 patients, and one plexiform neurofibroma. As mentioned above, this analysis used restriction enzymes that will not cut in 24

25 the presence of a methylated cytosine. If a cytosine of interest is unmethylated, these enzymes will cut the genomic DNA, and there will be no PCR product. If methylation is not complete at a site, there will be uncut band visible after separation by electorphoresis on a PAGE gel, but the concentration should appear less. As a control, TaqαI digest reactions were also set up, to ensure that the sites of interest in these samples did not contain any mutations that would prevent either enzyme from cutting, regardless of methylation status. Representative gels from these experiments can be seen in Figure 2-2. All four CpG sites are free of mutation in these samples, as they can be digested by TaqαI (left-most lane of three for each sample). Based on the presence of roughly equivalent amounts of uncut product in the methylation-sensitive reactions (right-most lane of three for each sample) when compared to the uncut control reaction (middle lane for each sample), it appears that the cytosine at each of the four CpG sites is methylated in all samples. The equivalent amount of product in the uncut control and methyl-sensitive reactions suggests that these sites are completely methylated, to the level of detection possible with this method. To further validate the extent of methylation at these four sites of interest, I chose the CpG site in exon23.2 for analysis by bisulfite sequencing. In DNA treated with sodium bisulfite, unmethylated cytosines are converted to uracils by deamination; uracil is not normally present in DNA (rather it is seen in place of thymine in RNA) and during PCR amplification, thymine takes the place of the uracil, leading to a C to T change at the unmethylated sites. The cloning of a sodium bisulfite treated PCR product into a vector allowed for the isolation of a single allele per bacterial colony. Sequencing was done to reveal the relative number of cytosines versus thymines at the CpG site of interest, allowing an estimate of the percent methylation at that site. Figure 2-3 shows the untreated sequence of exon23.2, as well as the chromatogram from the 25

26 sequencing of one of the clones. The sequence represented in the chromatogram is underlined in the untreated sequence for comparison. All 5 clones analyzed contained a cytosine at the CpG dinucleotide of interest (indicating that it is methylated), while all other cytosines in the fragment had been replaced by thymine. The agreement between the methylation specific restriction digest results and the bisulfite sequencing results for the exon23.2 site of interest suggests that the four germline mutation hotspots are all heavily methylated in somatic tissue. These sites are thus susceptible to CpG C to T transitions. Mutation Detection by PCR and Restriction Digest Somatic mutation detection analysis was carried out on DNA from 63 dermal neurofibromas and 83 plexiform neurofibromas (Table A-1). Each sample was PCR amplified with each of the four primer sets listed in Table 2-1 and then subjected to restriction digest by TaqαI. All 4 CpG sites were analyzed in 117 samples, and 1 to 3 of the CpG sites was analyzed in the remaining 29 (due to lack of PCR amplification by on or more primer pair). Across all the samples, 531 CpG dinucleotides were analyzed for C to T transitions. Figure 2-4 shows an example of the visualization of the digest products on an 8% PAGE, with uncut PCR product indicating presence of a mutation (normal sequence is digested into two smaller products). While this analysis was used to screen for somatic CpG C to T mutations, it also revealed two previously-identified constitutional mutations, one at the exon10a CpG site and the other at the exon23.2 (Figure 2-2) CpG site. These were the only mutations, germline or somatic, seen at any of the four CpG sites in the 146 neurofibromas, from 82 independent patients. Discussion The methylation of cytosine in a CpG dinucleotide is a common occurrence in the human genome. It is estimated that 70-80% of all CpG dinucleotides contain 5- methylcytosine (5mC) (Razin and Riggs, 1980). The percent of isolated CpG sites that are methylated is likely higher, 26

27 as this statistic includes all CpG dinucleotides, both isolated and those found in CpG islands, the latter of which account for the majority of non-methylated CpGs (reviewed by Bird, 2002). Based on these observations, it was expected that the four CpG sites studied here would be methylated in normal tissue. The results of my methylation status analysis confirm that all four of these sites are indeed predominantly, if not completely, methylated in normal Schwann cells and neurofibromas. The positive methylation status of these CpG sites makes them susceptible to C to T transitions due to the spontaneous deamination of 5meC. Since C to T transitions can occur without need for replication, this mutation mechanism is feasible in Schwann cells, which are typically quiescent unless stimulated to divide by injury, or occasional divisions to keep up with nerve growth. Interestingly, no such mutations were identified in 146 tumors analyzed at the CpG germline mutation hotspots. All four of these CpG sites have been previously identified as germline mutation hot spots with a 1-2% recurrence rate (arginine-to-stop), and the estimated combined germline mutation rate at these four sites is approximately 7% (they account for 7% of all NF1 germline mutations) based on our lab s data and published comprehensive studies from other labs. Thus, if the somatic mutation rate was equivalent, I would have expected to see approximately 37 total mutations (7% of 531 sites analyzed) among the four sites in the 146 tumors. Since no somatic mutations were identified, it appears that the rate of somatic CpG C to T transitions is very low in neurofibromas, (at least at these sites), despite the presence of a methylated cytosine. The detection of the two known germline mutations by my methods indicated that this apparent dearth of somatic CpG C to T mutations is not due to a faulty detection method. The level of mutation seen in my study is consistent with the low level reported in the NF1 mutation 27

28 literature. This also suggests that C to T transitions do not play a major role in somatic events in NF1 tumorigenesis. This is in contrast to the rate of CpG C to T mutations in other, non-nf1 tumors. The rate of somatic CpG C to T mutations in TP53 ranges from 25% in bladder cancer to nearly 50% in colon cancer (Jones et al., 1991; Greenblatt et al., 1994; Olivier et al., 2002). In a study of hereditary non-polyposis colorectal cancer, 30.7% (4/13) of somatic APC mutations were CpG C to T mutations (Huang et al., 2004). The differences in the rate of these mutations in neurofibromas compared to other tumor types may be due to differences in their ability to repair this type of mutation. As many cancers exhibit mutations in DNA repair pathways, it may also be that more malignant tumors are more susceptible to C to T transition mutations, (whereas neurofibromas are benign) due to reduced ability to repair the mismatches. The status of base excision repair in neurofibromas has not been analyzed, although it is known that cytogenetic rearrangements virtually never occur in dermal neurofibromas and less then half the time in plexiforms, so these tumors tend to have fairly good chromosomal stability (Wallace et al., 2000). However, NF2 schwannomas are also benign Schwann cell tumors that do not show chromosomal rearrangements, yet somatic CpG C to T mutations are very frequent. This may be pointing to basic differences in these tumor types for frequency of 5meC deamination and/or robustness of excision repair. Understanding the frequencies and mechanisms of CpG somatic mutation may help predict whether certain individuals or tumors are more at risk for these, or lead to a specific therapy for tumors containing such mutations. For example, there are several therapies that have shown potential in allowing translations through premature stop codons, including gentamicin and related compounds, and antisense oligonucleotides (reviewed by Ainsworth, 2005; Kulyte et al., 2005; Pinotti et al., 2006). 28

29 Table 2-1. Primers used in CpG C to T mutation screening. Exon Primer sequence (5-3 ) Product size (bp) Annealing temp ( C) 10a 5 ACGTAATTTTGTACTTTTTCTTCC 3 CAATAGAAAGGAGGTGAGATTC TGCTACTCTTTAGCTTCCTAC 3 CCTTAAAAGAAGACAATCAGCC TTTTAAGGAGTGATTTTTGTTATTTG 3 CCTTCTTTGATAAAGCATTCTTC TTCATCCTGTTTTAAGTCACACTTG 3 TTGCCTCCATTAGTTGGAAAATTG TaqαI digest band sizes (bp) Genomic DNA Digest with methylation-sensitive enzymes (BstBI for e10a, e41; AvaI for e22, e23.2) PCR with flanking primers PCR with flanking primers Digest with TaqαI Polyacrylamide gel Polyacrylamide gel Unmethylated C Methylated C not Me-sensitive No band Uncut band Cut band Figure 2-1. Methylation-sensitive restriction digest protocol. 29

30 m a b c d m m a b c d m TaqI BstBI AvaI E10a E22 m a b c d m m a b c d m TaqI BstBI BstBI E23.2 E41 Figure 2-2. Visualization of methylation-specific restriction digest analysis. a: normal human Schwann cell culture; b and c: non-nf1 patient blood; d: plexiform neurofibroma DNA. GTTAGAACCATCAGAGAGCCTTGAGGAAAACCAGCGGAACCTCCTTCAGATG ACTGAAAAGTTCTTCCATGCCATCATCAGTTCCTCCTCAGAATTCCCCCCTCA ACTTCGAAGTGTGTGCCACTGTTTATACCAGTTTATACCAGGTATGCTTACAG TTAGAGATTAC Figure 2-3. Results of bisulfite sequencing. The sequence given is the untreated sequence of exon23.2. Upon treatment with bisulfite followed by sequencing, unmethylated Cs are converted to Ts, methylated Cs are not converted. All Cs are in blue, the CpG site of interest is in red. The chromatogram shows the underlined region of sequence after bisulfite treatment. The only C not converted to T is in the CpG dinuleotide, indicating it is methylated. 30

31 U M * * Figure 2-4. TaqαI digests of E23.2 PCR products visualized on 8% PAGE. All samples are from plexiform neurofibromas. The sample indicated by the asterisk is from a patient with a known constitutional CpG mutation. C= control with known mutation at cut site; U= uncut sample; M=1 kb marker. 31

32 CHAPTER 3 ALTERNATIVE SPLICING OF EXON 23a AND mrna EDITING Introduction The alternative splicing of exons allows for the production of multiple different transcripts from a single gene. This usually affects the coding region in between invariant translation start and stop sites, but this phenomenon can also produce transcripts with different transcription or translation start and/or end sites. Alternative splicing is often regulated in a spatial, tissue-specific, and/or temporal manner, and its effects can generally be divided into five categories: effects on localization of the resulting protein, elimination of the resulting protein s function, changes in the resulting protein s function, creation of a new function of the protein, and effects at the RNA level, such as transcript stability or efficiency of translation (reviewed by Smith et al., 1989). While there are many examples of alternative splicing events, there is also evidence that disregulated alternative splicing plays a role in human cancers (Early et al., 1980; Nagoshi et al., 1988; Lee and Feinberg, 1997; Stimpfl et al., 2002; Adams et al., 2002). One array study identified 845 alternative splicing isoforms from throughout the genome that appear to be tumor associated (Wang et al., 2003). While no NF1 isoforms were included in this data set (which also did not include Schwann cell tumors), it is known that several common NF1 isoforms normally exist at at least a 10% level relative to the major isoform in the pertinent tissue. There are three alternatively spliced exons in NF1, as well as a short isoform that lacks exon11 through most of exon49 (Nishi et al., 1991; Suzuki et al., 1992; Gutmann et al., 1993a; reviewed by Skuse and Cappione, 1997). The first of these alternatively spliced exons to be identified was exon23a (Nishi et al., 1991). The inclusion of this exon produces an mrna containing 63 additional base pairs from intron23-2, leading to an in-frame insertion of 21 amino acids in neurofibromin, and is known as the Type II isoform. These additional amino acids are 32

33 within the NF1 GAP-related domain (GRD), and have been shown to change the hydrophobicity and secondary structure of this region of the protein (Nishi and Saya, 1991). While this isoform is still able to bind RAS and has GAP activity, Andersen et al. (1993) found that cells expressing the Type II isoform contained much more RAS-GTP (activated RAS) than cells expressing mostly the Type I isoform (lacking exon23a). This was shown to be due to ten-fold decreased GAP activity of the protein encoded by NF1 Type II mrna. It has also been shown that Type II neurofibromin does not associate with microtubules as Type I does (Gutmann et al., 1995). It has been suggested that in addition to altering or eliminating known functions of neurofibromin, the insertion of exon23a may introduce novel functions to the protein. Teinturier et al. (1992) found that exon23a showed sequence homology with vaccinia virus nucleoside triphosphotase I, and Andersen et al. (1993) point out similarities between the 21 amino acid insert and common nuclear localization signals. While the functional effects caused by the inclusion of exon23a have yet to be fully characterized, Costa et al (2001) found that exon23a knockout mice had an increased incidence of cognitive deficits, relative to the exon31 knockout mouse (which has no cognitive impairment). It also appears that this isoforms is developmentally significant, as there is a switch from the predominant expression of Type I to Type II neurofibromin through embryogenesis into post-natal life in the majority of tissues in rat, mouse and chick, as well as differentiating human cell types (Nishi et al., 1991; Baizer et al., 1993; Gutmann et al., 1994, 1995; Huynh et al., 1994; Mantani et al., 1994). Although there are some conflicting reports, the consensus is that the Type II transcript is typically the predominant transcript, present at levels greater than or equal to Type I in most adult tissues, with the exception of the adult central nervous system (Suzuki et al., 1992; Uchida et al., 1992; Teinturier et al., 1992; Baizer et al., 33

34 1993). Two of these studies also indicated that in various non-neurofibromatosis 1 (NF1) associated cancers the expression of Type II mrna is preferentially increased (Suzuki et al., 1992; Teinturier et al., 1992). Ogata et al. (2001) found that a particularly malignant breast cancer cell line (MDA-MB-231) does not express any type I NF1 mrna, further supporting a role of aberrant alternative splicing of exon23a in tumors. There are two other well-documented NF1 isoforms. One is expressed exclusively in the central nervous system. This isoform includes exon9br, which encodes 10 additional amino acids inserted at residue 420 of the protein (Danglot et al., 1995; Geist and Gutmann, 1996). The other includes exon48a, which is a muscle-specific 18 amino acid insertion near the C-terminal end of the neurofibromin molecule (Gutmann et al., 1993a). These were not included in this tumor study since they do not affect RAS-GAP activity, and do not appear to be involved in NF1 tumorigenesis. In addition to these well-studied isoforms, there have been several reports of multiple rare novel splice variants of NF1 mrna (Thomson and Wallace, 2002; Vandenbroucke et al., 2002a, 2002b). While some of these alternative splice events insert or delete intronic or exonic sequence, many of them involve exon skipping due to splicing at the same sites used in normal NF1 RNA (Thomson and Wallace, 2002; Vandenbroucke et al., 2002b). Some of these produce out-of-frame transcripts. Thomson and Wallace (2002) found that the conditions under which blood samples were drawn, or length of time stored before RNA was isolated, impacts this rare variant splicing profile of NF1, with the frequency of these increasing over time. However, relative ratios of the Type I to Type II mrna are not affected. Some of these novel isoforms have been shown to occur in a tissue specific manner, implying that they may be functionally significant despite relatively low levels (Vandenbroucke et al., 2002a, 2002b). It is clear that 34

35 these isoforms do exist in vivo, but further study is needed to determine the role of these rare NF1 isoforms in normal tissues and in NF1 phenotype development. Another form of post-transcriptional modification that has been shown to occur in NF1 mrna is RNA editing. Skuse et al. (1996) found that some NF1 mrna undergoes base modification editing at position An in-frame stop codon (R1306X) is created by the deamination of the transcript s cytosine at bp 3916 of the mrna to form a uracil. This nucleotide is within the region encoding the neurofibromin GRD. It is unclear what effect this premature stop codon may have on the cells that express the edited mrna, but if translated it would produce a truncated protein that would likely be degraded or lack full function. Additionally, editing can occur in mrna transcribed from either allele, potentially inactivating that allele regardless of NF1 gene mutation state. While this editing has been found at low levels (1.5-2%) in all cell types studied, there has been evidence that NF1 mrna editing occurs at somewhat higher levels in tumor cells (Skuse et al., 1996; Cappione et al., 1997; Mukhopadhyay et al., 2002). Cappione et al. (1997) observed a correlation between increased invasiveness of tumors with increased levels of mrna editing, with malignant tumors having higher levels of editing than plexiform neurofibromas, which in turn had higher levels of editing than dermal neurofibromas. Importantly, this trend at the NF1 gene does not appear to be due to an overall increase in general mrna editing in malignant cells (Cappione et al., 1997). Mukhopadhyay et al. (2002) evaluated RNA editing in malignant peripheral nerve sheath tumors (MPNSTs) and found that 76.5% of tumors examined exhibited low levels of mrna editing (0-2.5%), near the reliable detection limits of their assay. However, they also identified a distinct subpopulation of these tumors (23.5%) that exhibited mrna editing at higher levels (3-12%). Tumors that exhibited this higher, reproducible level of mrna editing at position 3916 also exhibited 35

36 increased levels of Type II mrna relative to Type I. The creation of this stop codon in some transcripts in cells expressing Type II mrna would further reduce neurofibromin function by leading to higher levels of activated RAS. These data were suggested to support a connection between RNA editing, alternative exon23a splicing, and tumorigenesis. The extent of these post-transcriptional modifications in NF1 tumorigenesis has yet to be validated by multiple independent groups. I examined alternative splicing of exon23a and mrna editing in our set of NF1 tumors and normal Schwann cells to test for a connection between these two post-transcriptional modifications. It is hoped that these studies will further clarify the roles of these modifications in NF1 tumorigenesis and provide potential new avenues of treatment, aimed at altering mrna expression and post-transcriptional modifications. Material and Methods Reverse-Transcription and PCR RNA was previously isolated from blood (8 non-nf1 patients, 7 NF1 patients), tumor (22 dermal neurofibromas, 21 plexiform neurofibromas, 6 MPNSTs), and culture samples (3 normal Schwann cell, 4 dermal tumor cultures, 9 plexiform tumor cultures, 2 immortalized plexiform neurofibroma cell lines) using the Trizol reagent and manufacturer s protocol (Invitrogen), and stored at -80 C. Reverse transcription reactions were carried out using Superscript II reverse transcription kit (Invitrogen) and random hexamer primers. The resulting cdnas were used in PCR reactions under standard conditions using primers in exon23.1 and exon24, flanking exon23a (CAT-H: 5 ATTGTGATCACATCCTCTGATTGG 3 ; CAT-I: 5 ATCTAAAATCCCTGCTTCATACGG 3 ). Two fragments were amplified, one each from Type I and Type II mrna (303 and 366 bp, respectively). The two isoforms were separated by electrophoresis on 8% native polyacrylamide gels and visualized by ethidium bromide staining. Based on visual observation of band intensities, the relative ratio of Type I to 36

37 Type II NF1 mrna was noted (I<<II, I<II, I II, I>II, etc.) and compared between tumor samples and controls, and between tumor types. Samples having only (or mostly) type II mrna were selected for cloning and sequencing to determine if RNA editing at C3916 was occurring in these tumors. Cloning and Sequence Analysis New RT-PCR products were generated from the cdna samples identified above, using primers E21c5 (5 CACAATGATGGGTGATCAAGG 3 ) and E23.1c3 (5 CATGTTGCCAATCAGAGGATG 3 ), spanning the 3916 edit site (377 bp fragment from Type II mrna). The products were ligated into a plasmid vector (Topo TA, Invitrogen) and transformed into chemically competent E. coli cells (One Shot TOP10, Invitrogen) using the manufacturer s protocol. The cells were plated on LB-ampicillin agar plates with IPTG and X-gal, and incubated at 37 C overnight. Bacterial colonies (clones) positive for the insert were identified by blue/white selection, followed by PCR amplification using the original primers. The PCR products from 30 positive clones were sequenced. Cycle sequencing was done using the ABI Big Dye 3 system and automated sequencers in the UF CEG sequencing core, with one of the PCR primers as a sequencing primer. Controls were used at all steps. Results Alternative Splicing Patterns I investigated the relative concentrations of Type I versus Type II mrna in fresh blood leukocytes, dermal neurofibromas, plexiform neurofibromas, MPNSTs, and multiple cell cultures including normal Schwann cells. The use of a single primer pair to amplify both isoforms helps preserve the ratio present in the original sample. Multiple samples of each tissue type were analyzed, to determine if there was a common splicing pattern for the given cell type, 37

38 and how that compared to the literature. A summary of the results of this analysis is given in Table 3-1, and a detailed list of samples analyzed can be found in Appendix B (Table B-1). In leukocytes from non-nf1 patients, Type I mrna was present at equal to or greater levels then Type II in 7/7 samples, with the majority showing Type I as the predominant transcript (5/7). Results from these samples are shown in Figure 3-1A. From left to right in the figure, lane 1 and 2 show more Type I than Type II, lanes 3 and 4 show approximately equal levels of the two isoforms, and lanes 5-7 again show more Type I than Type II transcript. In leukocytes from seven NF1 patients, there was more variability between samples. Two out of seven showed Type II as the predominant transcript, three out of seven had approximately equal amounts of the two isoforms, and the remaining two had more Type I transcript than Type II. Three cultures from normal human Schwann cells were analyzed. All three of these cultures contained Type II mrna as the main transcript, with one culture exhibiting a much higher level of Type II compared to Type I than the others. Both primary dermal neurofibroma tissue and Schwann cell cultures derived from dermal tumors were analyzed for their exon23a splicing patterns. Of the 21 primary dermal neurofibroma samples, there were two tumors from which corresponding cultures were also analyzed. An additional two dermal tumor-derived cultures were analyzed. The majority of primary dermal tumors contained predominantly Type II transcript (18/21), with 7 having very low levels of Type I in comparison. Two dermal tumor samples, UF80T32 and UF505T4, had Type I as the main transcript. One primary dermal sample, UF80T2, contained no Type I transcript, and was selected for cloning and sequencing to test for RNA editing. All 4 dermal tumor-derived cultures contained predominantly Type II transcript. Figure 3-1B shows experimental results from some representative dermal samples. Lane 1 is a sample with more 38

39 Type I mrna, lanes 2-5 and 7 are samples with predominantly Type II transcript, and lane 6 is UF80T2 (which has no detectable Type I transcript). Twenty-five plexiform neurofibromas and nine cultures derived from plexiform tumors were also analyzed for exon23a splicing. Of the nine cultures analyzed, the corresponding primary tumor was also analyzed for four. The majority of primary plexiform tumors (23/25) contained predominantly Type II transcript, three of which had very little Type I in comparison, and two having only trace amounts of Type I. Of the remaining two tumors, one had about equal levels of Type I and Type II mrna, and the other had Type I as the predominant transcript. The majority of plexiform tumor-derived cultures (8/9) also had Type II as the predominant transcript, with 3 having barely detectable levels of Type I transcript. The remaining culture, UF469Tc, contained only Type II transcript and was also selected for RNA editing analysis. Figure 3-1B shows the experimental results from representative primary plexiform tumors, and one plexiform-derived culture. Lanes 8-10, and lane 12 show samples with predominantly Type II transcript, and lane 11 is UF469Tc, with only Type II mrna. Finally, 6 MPNSTs were analyzed. All 6 had Type II mrna as the predominant transcript, with 3 showing much lower levels of Type I in comparison. Figure 3-1C shows the experimental results for representative MPNSTs. Lanes 1 and 2 show samples with much more Type II then Type I transcript, while lanes 3 and 4 show samples with predominantly Type II mrna. One MPNST, SNF94.3, had previously been analyzed for exon23a splicing levels, and was found to contain only Type II transcript. This sample was also chosen for RNA editing analysis. Analysis of RNA Editing Three samples were selected for RNA editing analysis based on their exon23a splicing patterns. All three samples contained only Type II transcript, based on polyacrylamide gel 39

40 electrophoresis, but they were all from different tissue types. UF80T2 is a primary dermal neurofibroma, UF469Tc is a plexiform neurofibroma-derived Schwann cell culture, and SNF94.3 is an MPNST. Plasmid vectors containing individual cdna fragments were cloned in E. coli and then PCR amplified for sequencing. This allowed for the analysis of the sequence of a single cdna molecule at a time, to detect the number of transcripts that were undergoing RNA editing at C3916. Thirty-five (UF469Tc) to Forty (SNF94.3 and UF80T2) clones were analyzed for each sample of interest. This number was chosen based on the levels of RNA editing previously detected in MPNSTs lacking Type I transcript (3-12%) (Mukhopadhyay et al., 2002). Even at the lowest levels previously seen, at least one in forty clones would be expected to contain an edited cdna fragment. No RNA editing was seen in any of these tumor samples, despite previous reports of increased levels of RNA editing in tumors containing predominantly Type II transcript. Discussion The relative ratios of Type I to Type II NF1 mrna in a given tissue sample can be determined using a single set of PCR primers to amplify the fragment of interest. The alternative splicing pattern of exon23a has previously been studied for many embryonic and adult solid tissues in mouse and human, but the levels of alternative splicing seen in blood leukocytes were not well established. I observed that the ratio of Type I to Type II NF1 transcript in non-nf1 blood leukocytes is similar to that reported in the adult human central nervous system, with all the samples (7/7) showing Type I transcript at equal or greater levels compared to Type II, with 5 of those 7 having Type I as the predominant transcript. This is in contrast to most other human and mouse postembryonic tissues studied, where Type II predominates. The deviation from this pattern in leukocytes is interesting. It is known that some alleles with NF1 mutations have decreased mrna levels, presumably due to nonsense-mediated decay (Colman et al., 1993; 40

41 Hoffmeyer et al., 1995; Pros et al., 2006). However, there should be no such mechanism operating in normal leukocytes, and it is known that this alternative splicing affects both alleles (Thomson and Wallace, 2002). Thus, there may be functional reasons in leukocytes and the central nervous system for a relative lack of Type II transcript, and the presence of an NF1 gene mutation shouldn t affect relative ratios of Type I to Type II transcript. Yet in blood leukocytes from NF1 patients, there is more variation in the relative ratio of Type I to Type II transcript observed. Only 5/7 samples showed relative ratios similar to those seen in the majority of non-nf1 bloods, with the majority of those 5 having equal levels of the two transcripts. It appears that there may be a trend toward inclusion of exon23a in the mrna from NF1 patient leukocytes, compared to leukocytes from non-nf1 patients. To determine if the level of exon23a inclusion varied based on tumor type, I analyzed the relative ratio of Type I versus Type II mrna in both dermal and plexiform neurofibroma samples. There was no significant difference in these relative ratios between the two types of neurofibromas. The majority of both dermal (18/21) and plexiform (23/25) neurofibromas had Type II as the predominant transcript. This is not unexpected as cultures of normal human Schwann cells, the cell type from which neurofibromas are clonally-derived, all contained more Type II mrna then Type I. Observation of a relative (but not dramatic) predominance of Type II transcript suggests a relatively reduced RAS-GAP activity in those cells, which would be inferred for most normal adult tissues, including, as I have shown here, Schwann cells, based on their Type I to Type II ratio. However, many of the NF1 tumors had a profile of Type II transcript at much higher levels relative to Type I, with several tumors, of each variety of neurofibroma, showing near or complete loss of the Type I transcript. This could result in sufficiently less RAS-GAP activity relative to the native Schwann cell, which could contribute to 41

42 tumorigenesis. If true, then a mechanism to shift the ratio back toward more equal levels of the two transcripts could potentially decrease tumorigenic potential. Previous reports had indicated that sample collection conditions and other environmental factors may influence the level of alternative splicing in NF1 mrna (Thomson and Wallace, 2002). To determine if culturing condition altered the level of exon23a inclusion, I examined the relative ratios of Type I to Type II NF1 mrna in several primary tumor and corresponding tumor Schwann cell culture pairs. Six such pairs were analyzed: two were dermal neurofibromas and their corresponding cultured Schwann cells, and four were plexiform neurofibromas and their corresponding Schwann cell cultures. All 6 primary tumors analyzed contained predominantly Type II mrna, and there was no obvious change in the relative ratio of the two isoforms compared to the corresponding cultures. The result of one of these comparisons is shown in Figure 3-2 (plexiform neurofibromas). These results suggest that there is an inherent Type I:Type II control mechanism in these tumor Schwann cells that is not susceptible to influences of tissue culture, and it appears that the relative ratio of Type I to Type II mrna seen in a tumor cell culture can be a good estimate of the ratio seen in the primary tumor. It was previously reported that MPNSTs containing little or no Type I transcript may exhibit RNA editing (Mukhopadhyay et al., 2002). All six of the MPNSTs analyzed here showed Type II as the predominant transcript and one (SNF94.3) that had been previously analyzed showed complete loss of the Type I transcript by ethidium bromide visualization. I analyzed this MPNST, as well as one dermal neurofibroma and one plexiform neurofibroma-derived culture, both of which also showed loss of the Type I transcript, for the presence of RNA editing. Thirty-five to forty clones from each sample were sequenced to check for RNA editing at C3916, but this edit was not seen in any of the three samples with little to no 42

43 Type I transcript. Mukhopadhyay et al. (2002) reported levels of RNA editing in their MPNSTs of interest at 3-12%. If RNA editing occurred to a similar extent in all tumors showing a near or complete loss of the Type I transcript, I would have expected to see 1-4 clones out of 40 undergoing RNA editing. The absence of editing in my samples could have been due simply to chance. However, it could imply that any connection seen between increased levels of Type II mrna and RNA editing is not universal, even in MPNSTs of this type, or that the amount of such editing is quite low, in which case its functional significance would seem minimal. While the creation of a premature stop codon in NF1 mrna by RNA editing has a clear negative effect on neurofibromin function, the effect of having predominantly Type II transcript may also have a negative impact, as the protein encoded by this transcript has a reduced RAS-GAP activity. The first MPNST that I studied for RNA editing has one of the germline NF1 microdeletions, but no somatic mutation has been found in the remaining allele despite sequencing the entire mrna open reading frame and the 3 UTR. This raises the interesting possibility that the alternative splicing in this tumor substitutes as a mutation, producing a hypomorphic allele. The cells thus would lack sufficient NF1 RAS-GAP activity to prevent tumorigenesis. This suggests that there may be a threshold of neurofibromin activity that keeps a cell from becoming tumorigenic. This is an important notion that could have relevance for future therapies. Alternatively, there could be one of several epigenetic changes (other than RNA editing) that could constitute the second hit in this MPNST, and possibly other NF1 tumors. These include histone modifications, changes in microrna effects, and changes in downstream regulatory effects (Schmegner et al., 2005; Ling et al., 2006; Martinez and Schackert, 2007; Shelton et al., 2008; Bartels and Tsongalis, 2009). The latter is evidenced by studies by Hawes et al. (2007) that found that different mouse strains have different levels of Nf1 expression. This 43

44 implies that different background levels of transcription factors can have an effect on NF1 expression levels (Zhu et al., 2004). The observation of different Type I to Type II ratios in two dermal neurofibromas from the same individual (UF80T2 and UF80T32; UF505T4 and UF505T7) suggests that this ratio is specific to each tumor and is not heavily controlled by systemic factors. This is consistent with the fact that each neurofibroma has a different NF1 somatic mutation (and possibly other genetic or epigenetic alterations) and is therefore independent. 44

45 Table 3-1. Summary of alternative splicing of exon23a seen in various sample types examined. Total mrna levels Cell Type sample # No Type I Type I < II Type I II Type I > II Normal Schwann cell culture 3 3 (1+) Non-NF1 patient blood Non-NF1 patient fibroblasts 1 1 NF1 patient blood Primary dermal tumor (5+, 2*) 2 Dermal tumor cell culture 4 4 (2+) Primary plexiform tumor (3+, 2*) 1 1 Plexiform tumor cell culture (1+, 3*) Immortalized plexiform cell lines 2 2 MPNSTs (3+) +: Much more Type II; *: Barely detectable Type I in comparison to Type II. 45

46 A B C Figure 3-1. Representative gels showing relative concentrations of Type I v Type II mrna in various tissue types studied. Upper band represents Type II mrna, lower band Type I. A) leukocytes from 7 non-nf1 patients. B) Primary tissue samples from dermal (lanes 1-7) and plexiform (lanes 8-12) tumors. C) Cultured (left two lanes) and primary tissue (right two lanes) from MPNSts. C T C T Figure 3-2. Comparison of alternative splicing of exon23a in primary plexiform tumors v cultured Schwann cells from the same tumors. Tumor culture sample is loaded first, corresponding primary tumor sample in the lane to the right. T= primary tumor, C= tumor culture. 46